RAG (Retrieval-Augmented Generation)

What does this mean?

RAG combines information retrieval with text generation. Instead of relying only on training data, the system first searches a knowledge base and uses the retrieved documents as the basis for its response.

How it works

1. User asks a question → 2) System searches relevant documents in a vector database → 3) Retrieved texts are passed as context to the LLM → 4) LLM generates a response based on the documents.

Example

An internal AI assistant that answers questions about company policies by finding and citing the relevant policy documents.

Why it matters

RAG is the standard approach for equipping LLMs with company-specific knowledge — without expensive fine-tuning.